AttributeError: ‘NumPy NdArray’ Object Has No Attribute ‘IsIn’ | Fast Fix

Guide / By

Use np.isin or np.in1d with arrays; ndarray has no .isin method—use Series.isin only on pandas objects.

This error pops up when array code calls a method that doesn’t exist on a NumPy ndarray. The root is simple: NumPy exposes isin as a function (np.isin(element, test_elements)), not as an array method, while pandas exposes .isin() as a Series/DataFrame method. A second cause is an older NumPy version that lacks np.isin. You’ll fix it fast once you match the API to the object in use and, if needed, update your environment.

What The Error Really Means

Python raises an attribute error when code tries to access a property or method that isn’t defined for that object. In this case, the object is a numpy.ndarray, and the missing attribute is a method named IsIn or isin. NumPy arrays do not implement a callable .isin method, so arr.isin(...) fails. The correct NumPy call is the top-level function np.isin(arr, values), which returns a boolean array. Pandas objects behave differently: Series and DataFrame do expose .isin, and that often causes muscle memory mistakes when jumping between libraries.

Quick check: glance at your variable type. If it prints as type(x) is numpy.ndarray, you should use np.isin or np.in1d. If it prints as pandas.Series or pandas.DataFrame, you can call .isin as a method.

AttributeError: ‘NumPy NdArray’ Object Has No Attribute ‘IsIn’

Case sensitivity also bites here. The canonical function is lowercase: np.isin. Spelling variants such as IsIn or isIn will fail. A second wrinkle: very old NumPy versions didn’t ship np.isin at all; projects pinned to NumPy < 1.13 can trigger a module-level attribute error for np.isin, not just a method call on an array.

  • Confirm the API — Use np.isin for arrays, and .isin on pandas Series/DataFrame.
  • Use lowercase — Call np.isin(...), not np.IsIn or arr.isIn.
  • Check NumPy version — np.isin exists in modern NumPy; if missing, fall back to np.in1d or upgrade.

Fix ‘Ndarray Has No Attribute Isin’ By Context (With Safe Patterns)

Pick the pattern that matches your object and goal. Each fix returns a boolean mask you can use for indexing or conversion.

Use NumPy With NumPy Arrays

# element-wise membership test with broadcasting
import numpy as np

arr = np.array([[1, 2, 3],
                [4, 5, 6]])
keep = np.array([2, 5, 99])

mask = np.isin(arr, keep)     # boolean array, same shape as arr
filtered = arr[mask]          # 1D view of matching elements
# For row/column level filtering, reduce along an axis:
rows_with_keep = np.isin(arr, keep).any(axis=1)
arr_rows = arr[rows_with_keep]

Fallback For Older NumPy

# if np.isin is unavailable, use np.in1d on flattened inputs
import numpy as np

arr = np.array([1, 2, 3, 4, 5])
keep = np.array([2, 5])

mask = np.in1d(arr, keep)     # 1D only
filtered = arr[mask]

Note: np.in1d works on 1D input. For 2D+, flatten before the check and reshape the mask back if you need the original geometry.

Use Pandas Method On Pandas Objects

import pandas as pd

s = pd.Series(["a", "b", "c", "a"])
mask = s.isin(["a", "c"])  # Series method
result = s[mask]

df = pd.DataFrame({"x": [1, 2, 3], "y": [4, 5, 6]})
mask_df = df.isin({"x": [2, 3]})  # DataFrame method
rows = df[mask_df.any(axis=1)]

Convert Between Pandas And NumPy Cleanly

# pandas -> numpy
arr = s.to_numpy()
mask = np.isin(arr, [1, 3])

# numpy -> pandas
import numpy as np, pandas as pd
arr = np.array([10, 20, 30, 40])
mask = np.isin(arr, [20, 40])
s = pd.Series(arr)
filtered = s[mask]

Version And API Checks Before You Change Code

Tooling drift causes many AttributeErrors. A short round of checks prevents noisy patches.

  • Print versions — import numpy as np, pandas as pd; print(np.__version__, pd.__version__).
  • Confirm function is present — hasattr(np, "isin") should return True on modern NumPy.
  • Pin compatible versions — keep a requirements.txt or pyproject.toml with versions that ship the APIs you call.
  • Avoid shadowing — a file named numpy.py or a variable named numpy/np masks the real package. Rename local files and variables to avoid import confusion.
Context Use Example
NumPy array, any shape np.isin(element, test_elements) np.isin(arr, [2, 5])
NumPy array, 1D only np.in1d(a, b) np.in1d(a, b)
Pandas Series or DataFrame obj.isin(values) df.isin({"col": [1, 2]})

Edge Cases That Trigger Confusion

Some inputs change behavior or speed. Here’s what to watch for.

  • Object dtype — mixed types or Python objects can slow np.isin/np.in1d and change equality rules. Clean or cast data when you can.
  • Strings vs lists — passing a single string as values to pandas .isin raises a type issue; wrap it in a list. For NumPy, np.isin(arr, "abc") will compare each element to the string "abc" and rarely matches; pass ["a","b","c"] only when that’s your intent.
  • Broadcasting — np.isin broadcasts element against test_elements. If you expect row-wise logic, reduce with .any(axis=1) or .all(axis=1) after the call.
  • Set-like tests — for large whitelist/blacklist arrays, de-duplicate test_elements first or set assume_unique=True when using compatible NumPy versions to speed things up.
  • Memory spikes — very large 2D masks can be heavy. If you only need to test one column from a 2D array, slice first, then call np.isin.

Performance Notes And Practical Patterns

Both libraries can be fast, but the faster option depends on shapes and dtypes. Large numeric sets often run well in NumPy. Table-shaped data that you filter by column usually reads cleaner in pandas with .isin. Always measure with real inputs.

  • Trim inputs — drop duplicates in test_elements before the check.
  • Choose 1D where possible — slice to one column, run the membership test, then use the mask to subset rows.
  • Prefer vectorized code — avoid Python loops that build masks one item at a time.
# fast row filter: keep rows where col "x" appears in a list
import numpy as np

arr = np.array([[10,  1],
                [20,  2],
                [30,  3],
                [40,  4]])
keep = [20, 40]

rows = np.isin(arr[:, 0], keep)   # build mask from first column
out = arr[rows]

Common Mistakes And How To Rewrite Them

  • Calling a method on ndarray — arr.isin([1,2]) ➜ np.isin(arr, [1,2]).
  • Using camel case — arr.isIn([1,2]) ➜ np.isin(arr, [1,2]).
  • Confusing pandas with NumPy — np_array.isin(vals) ➜ wrap as pd.Series(np_array).isin(vals) or stay in NumPy with np.isin.
  • Old NumPy — np.isin missing ➜ use np.in1d or upgrade NumPy.
  • Wrong axis logic — need row selection ➜ np.isin(arr, vals).any(axis=1), then index rows.

Drop-In Recipes You Can Paste

Keep Rows Where A Column Value Appears In A List

import numpy as np

def keep_rows_by_column(arr, col, values):
    mask = np.isin(arr[:, col], values)
    return arr[mask]

# usage
result = keep_rows_by_column(np.array([[1, 10], [2, 20], [3, 30]]), col=0, values=[1, 3])

Blacklist Values (Invert)

import numpy as np

arr = np.array([1, 2, 3, 4, 5])
ban = [2, 5]

mask = np.isin(arr, ban, invert=True)
clean = arr[mask]   # [1, 3, 4]

2D Mask With Restore Of Shape Using in1d

import numpy as np

arr = np.array([[1, 2, 3], [4, 5, 6]])
vals = [2, 5]

flat_mask = np.in1d(arr.ravel(), vals)
mask = flat_mask.reshape(arr.shape)
picked = arr[mask]

Why This Fix Works

The NumPy docs state that np.isin calculates membership and returns a boolean array with the same shape as the input, while np.in1d provides the same check for 1D arrays. The pandas reference shows that Series.isin and DataFrame.isin return boolean masks aligned to the original index and columns. Put together, the cure is to call the API that matches the object type and version in use.

The phrase attributeerror: ‘numpy ndarray’ object has no attribute ‘isin’ appears most often when pandas habits leak into array-only code. Once you swap to np.isin or switch the object to a pandas Series, the program moves on. If your logs show the exact message attributeerror: ‘numpy ndarray’ object has no attribute ‘isin’, scan for a stray .isin on an array or a missing import.

References